Hierarchy effects in copular constructions: The PCC corner of German*
Jessica Coon1 , Stefan Keine2 , & Michael Wagner1
1 McGill
1.
University, 2 University of Southern California
Introduction
This paper presents a new generalization about agreement in German copula constructions,
and proposes an analysis that ties it to other well-established phenomena. Specifically,
we demonstrate that German shows hierarchy effects similar to those observed in other
languages: Person Case Constraing (PCC) effects (e.g. in Romance, Basque), inverse constructions (e.g. in Algonquian), Agent Focus (e.g. in Mayan), and certain dative–nominative
patterns (e.g. in Icelandic). Specifically, we propose that what German copula constructions
have in common with these environments is that there there are multiple accessible DPs
in the domain of a single agreement probe (see e.g. Béjar & Rezac 2003, 2009, Anagnostopoulou 2005, Adger & Harbour 2007, Nevins 2007, Preminger 2014). We develop a
Multiple Agree account (Hiraiwa 2001, Nevins 2007) which both derives apparent hierarchy
effects from independent principles, and provides a new explanation for the apparent absence
of “Number Case Constraint” (Num-CC) effects (cf. Nevins 2011).
2.
Hierarchy effects in copular constructions
Heycock (2012) has observed that in German specificational copula constructions in which
one DP is a 1st or 2nd person and the other is 3rd, the copula invariably agrees with the
1st/2nd person DP. This pattern stands in striking contrast to English, where verb agreement
is always controlled by the DP which is linearly first. We will use the term ‘DP1 ’ to refer to
the linearly first DP in a copular construction, and ‘DP2 ’ for the linearly second one.
(1)
Du
bist/*ist das Problem.
you.NOM are/*is the problem
‘You are/*is the problem.’
* Many
(2)
Das Problem bist/*ist du.
the problem are/*is you.NOM
cf. English: ‘The problem is/*are
you.’
thanks to Megan Jezewski for programming the experiment. For helpful feedback and discussion,
we would like to thank Boris Harizanov, Laura Kalin, Omer Preminger, and audiences at NELS and MIT.
Coon and Wagner gratefully acknowledge funding from the Canada Research Chair program.
Coon, Keine, & Wagner
Assuming that 1st and 2nd person DPs are featurally more marked than 3rd person
ones, one possible analysis is that agreement in these constructions targets the most marked
DP in the agreement domain, regardless of its structural position. Another option is that if
the predicate DP is more marked than the subject DP of the predication, it moves past the
subject to SpecTP, and then controls agreement (cf. Heycock 2012). Both accounts share that
the copula agrees with whichever DP is more marked, and we refer to them as ‘maximize
agreement accounts’. Another account (Heggie 1988, i.a.) is that the copula consistently
agrees with the underlying subject of the predication. In sentences such as (2), the subject
of the sentence is du and the predicate DP das Problem is moved over it to the clause-initial
position, as in (3) (see Béjar & Kahnemuyipour to appear, henceforth B&K, for a review).
(3)
Das Problem bist [ du bist das Problem ]
An immediate question that arises under this latter view is why in (2) the less marked DP
cannot function as the subject and control agreement. On a maximize agreement account,
this underlying configuration would result in agreement with the more marked predicate
DP (either through inversion or directly). We propose, by contrast, that copula agreement
invariably targets the subject of the predication, and that a structure with das Problem as the
underlying subject and du as the predicate is in fact ineffable. More generally, we propose
that the subject may only be 3rd person if the other DP is also 3rd person. German copula
constructions thus exhibit hierarchy effects akin to the (weak) PCC.
To support this claim, we investigate a particular kind of copula sentence, which Heycock
(2012) calls ‘assumed identity sentences’. In these constructions, the subject DP is assigned
the role or place of the predicate DP (e.g., when assigning roles in a play). An example is
provided in (4a). The utility of these constructions for our purposes lies in the fact that they
are sufficiently semantically asymmetric to reveal the underlying subject–predicate relation.
While the sentences in (1) and (2) are truth-conditionally equivalent, the role assignment I
am him is truth-conditionally different from the role assignment He is me.
We show that in exactly these constructions, hierarchy effects appear. Configurations
in which DP1 is a 1st or 2nd person Part(icipant) and DP2 is 3rd person (Part > 3) are
grammatical, as in (4a). The reverse (3 > Part) is ungrammatical, as in (4b). We also observe
a number hierarchy effect: PL > SG is possible (5a), whereas SG > PL is not (5b).
(4)
Person hierarchy: Part > 3
a.
Ich
bin er.
I.NOM am he.NOM
b. *Er
ist ich.
he.NOM is I.NOM
(5)
Number hierarchy: PL > SG
a.
Sie
sind er.
they.NOM are he.NOM
b. *Er
ist sie.
he.NOM is they.NOM
We report in section 3 the results of a sentence-rating experiment that supports the
conclusion that these constructions are subject to the person hierarchy in (4) and the number
hierarchy in (5). German copula constructions are ineffable if DP2 is higher than DP1 on
Hierarchy effects in copular constructions
either of these hierarchies. This stands in contrast with what is reported in Heycock 2012,
also discussed in B&K. In this respect, German copular constructions differ from English,
where all person and number combinations are well-formed.
The person hierarchy (4) is strongly reminiscent of the weak PCC. After establishing the
empirical generalization in section 3, we propose in section 4 an analysis that likens these
hierarchy effects to the weak PCC, drawing in particular on Multiple Agree analyses of
this phenomenon. We then propose, following Heycock (2012), that the difference between
German and English with respect to these two hierarchies follows from the fact that in
German both DP1 and DP2 are nominative and hence visible for agreement, whereas in
English, DP2 is accusative and unavailable (see Bobaljik 2008 and work cited there on
agreement accessibility). We then address a question that emerges from our account, namely
why number hierarchy effects arise in German copular constructions, but are absent in
ditransitive constructions. We offer a new solution that derives the absence of a “Number
Case Constraint” (Nevins 2011) from independently-motivated ingredients.
3.
Experiment
3.1
Design
We conducted a sentence-rating experiment for both English and German. The experiment
used “assumed identity sentences” like (6a) and (6b) and systematically manipulated the
person and number specification of the two DPs. To elicit ratings for the assumed identity
interpretation, a role-playing background was provided in which specific roles were assigned.
Each trial in the experiment consisted of rating one assignment.
(6)
a.
(pointing at you, then at your friend John)
You are him.
b.
(zeigt auf dich, dann auf deinen Freund Karl)
Du bist er.
Participants were asked to rate each sentence on a 6-point scale with ‘1’ being completely
unacceptable and ‘6’ being completely acceptable. As a control condition, the experiment
also included uncontroversially ungrammatical sentences in which the verb agreement is
inconsistent with either argument (*You am him; *Du bin er). 23 participants took part in
the English experiment. The German experiment had 15 participants.
One unusual aspect of the experiment is that it is impossible to lexically vary the target
structures (e.g., I am you). Given that there is only one possible lexicalization of each
condition, we did not manipulate item as a random effect. All participants saw the same
sentences, but the order of presentation was randomized.
3.2
Results
The by-condition means for the person hierarchy from (4) above, averaged over number, are
given in the form of boxplots in (7). In (8) we provide the means for the number hierarchy
Coon, Keine, & Wagner
from (5), averaged over person. The number above each boxplot represents the condition
mean. The column ‘Plateau’ refers to configurations in which both DPs instantiate the same
value of a feature, i.e., SG>SG, PL>PL, Part>Part, or 3>3.
(7)
Effects of person hierarchy
(8)
Effects of number hierarchy
We analyzed the data with cumulative link mixed regression models, using the R package
Ordinal (Christensen 2015). We fitted a model that predicted rating responses from the
predictors (i) person hierarchy (Part>3 vs. 3>Part vs. Plateau), (ii) number hierarchy (SG>PL
vs. PL>SG vs. Plateau), (iii) language (English vs. German), (iv) the interaction between
person and language, and (v) the interaction between number and language. The factor
language was sum-coded (English: –.5; German: .5). The factors person and number were
Helmert coded. In each case, the first comparison contrasted plateau configurations with
non-plateau ones (plateau: −2/3, non-plateau: 1/3). The second contrast compared the two
non-plateau configurations to each other (for person Part>3: .5, 3>Part: –.5, plateau: 0; for
number PL>SG: –.5, SG>PL: .5, plateau: 0). The models comprised the full random-effects
structure, namely, random intercepts and slopes by participants for all fixed effects and the
correlations between them.
The coefficients of this model are provided in (9), where ‘plt’ abbreviates ‘plateau’.
The model revealed significant main effects of the person and number hierarchy: Part > 3
configurations are rated higher than 3 > Part configurations and PL > SG structures are rated
as better than SG > PL. Crucially, there was an interaction between these hierarchies and
language such that the effect of the two hierarchies was greater in German than in English.
In order to investigate these interactions more closely in the individual languages, we
fitted a second model that nested the predictors person hierarchy and number hierarchy under
the levels of the factor language. The full random-effects structure of the original model
was preserved. The coefficients for this model are provided in (10). The model detected
that in German, 3 > Part configurations are degraded relative to Part > 3 configurations
and that SG > PL is worse than PL > SG. Interestingly, we also found that English shared
with German the preference for Part > 3 over 3 > Part. Notably, however, this effect was
Hierarchy effects in copular constructions
significantly smaller than in German. This effect may reflect a pragmatic preference for
encoding a participant argument rather than a 3rd person argument as the subject, given
the inherent availability (and topicality) of the participants of the discourse. In English,
number plateau configurations were also rated higher than non-plateau configurations. This
plausibly reflects the marked pragmatic status of person mismatches between the two DPs
in the provided role-playing scenario, not properties of narrow syntax.
(9)
(10)
Full model
Nested model
β̂ (SE)
Language
Person
Plt.vs.Non-plt
Part>3.vs.3>Part
0.38 (0.23)
−1.03 (0.23) ∗∗∗
Number
Plt.vs.Non-plt
PL>SG.vs.SG>PL
−1.06 (0.16) ∗∗∗
−0.83 (0.21) ∗∗∗
2.17 (0.64) ∗∗∗
Language
Person:Language
Plt.vs.Non-plt:Lang
Part>3.vs.3>Part:Lang
0.43 (0.44)
−1.12 (0.41) ∗∗
Number:Language
Plt.vs.Non-plt:Lang
PL>SG.vs.SG>PL:Lang
−0.80 (0.30) ∗∗
−1.69 (0.40) ∗∗∗
∗∗∗ p
3.3
β̂ (SE)
< 0.001, ∗∗ p < 0.01, ∗ p < 0.05
German
Person
Plt.vs.Non-plt
Part>3.vs.3>Part
Number
Plt.vs.Non-plt
PL>SG.vs.SG>PL
English
Person
Plt.vs.Non-plt
Part>3.vs.3>Part
Number
Plt.vs.Non-plt
PL>SG.vs.SG>PL
∗∗∗ p
2.17 (0.64) ∗∗∗
0.60 (0.36)
−1.59 (0.37) ∗∗∗
−1.46 (0.27) ∗∗∗
−1.67 (0.34) ∗∗∗
0.17 (0.26)
−0.46 (0.22) ∗
−0.66 (0.17) ∗∗∗
0.02 (0.23)
< 0.001, ∗∗ p < 0.01, ∗ p < 0.05
Discussion
The results show that assumed identity copula constructions are subject to the person
hierarchy (4) and the number hierarchy (5) in German. The interaction with language (in
the full model) shows that the size of the effects are significantly greater in German than in
English and hence that they go beyond mere effects of pragmatics in German.
Our account, developed section 4, attributes hierarchy effects to the interaction of
a φ-probe with two accessible (here, nominative) DPs. Under this analysis, all copula
constructions in languages like German are in fact subject to hierarchy effects. But this is
obscured by the possibility of surface inversion in copula constructions other than assumed
identity copulas (see (3) above), where construing the argument higher on the hierarchy as
the subject leads to a proposition that is truth-conditionally equivalent.
We should note, however, that while the configurations that violated the hierarchies received reliably lower ratings in German, they still received a relatively high rating compared
to our ungrammatical controls (4.7 in (7) and 4.4 in (8), vs. 1.4 for the controls). One reason
for this difference may be that in our control cases, agreement is inconsistent with either DP,
an error that is easily detectable, while in our test sentences verb agreement is consistent with
one of them. However, this is still not a sufficient explanation of their relative acceptability:
In copulas other than the assumed identity cases which are semantically symmetric, agreeing
Coon, Keine, & Wagner
with the argument lower on the hierarchy is clearly unacceptable (*Das Problem ist du in
(2)). While we did not test such cases in our experiment, informal elicitation leaves little
doubt that they are worse than our hierarchy-violating assumed-identity sentences. This is
compatible with the maximize agreement account, but unexpected under our proposal. We
believe that the ratings of the assumed-identity sentences might be inflated because there is
simply no other way of expressing the intended meaning in these cases, other than resorting
to a full accusative-assigning predicate.
4.
Person, number, and the PCC
4.1
Proposal
We propose that the hierarchy effects in German copulas arise due to the same confluence of
factors which have been proposed to cause hierarchy effects in a variety of other constructions cross-linguistically: two accessible DPs in the domain of a single φ-probe. Like other
recent work in this domain (e.g. Béjar & Rezac 2003, 2009, Anagnostopoulou 2005, Adger
& Harbour 2007, Nevins 2007, Preminger 2014), our account derives hierarchy effects from
independent principles; the hierarchy itself has no independent status in the grammar.
We focus first on the person hierarchy effects. Note that the generalization governing
the distribution of person features in German copula constructions is the same as the one
governing the combinations of direct and indirect object clitics in what is known as the
“weak PCC” in languages like Catalan (see Nevins 2007 and work cited there). The pattern
is shown in (11). In PCC configurations, DP1 is the indirect object and DP2 is the direct
object. In German copula constructions, DP1 is the subject and DP2 is the predicate.
(11)
Generalized weak PCC
In DP1 > DP2 , if DP2 is 1ST /2ND -person, then DP1 is 1ST /2ND -person.
We propose an account of German copula constructions which draws on recent literature
on PCC effects. First, we adopt the common assumption that 1st and 2nd person DPs bear
the feature [+Part(icipant)], whereas 3rd person DPs are [–Part(icipant)]. The feature [+Part]
is marked. We further adopt a version of the Person Licensing Condition (PLC) in (12),
which requires [+Part] to be licensed by entering into an Agree relationship with a probe
(Preminger 2016, adapted from Béjar & Rezac 2003).
(12)
Person Licensing Condition: A [+Part] feature on a DP that is a viable agreement
target (as far as case, etc. is concerned), and for which there is a clausemate person
probe, must participate in a valuation relation.
Finally, we assume that a single head may agree with more than one DP (Multiple
Agree), as long as the resulting Agree dependency satisfies Contiguous Agree (13); see
Hiraiwa 2001, Anagnostopoulou 2005, Nevins 2007.
Hierarchy effects in copular constructions
(13)
Contiguous Agree
Agree in a marked feature across an unmarked intervener is prohibited.
For the case at hand, Contiguous Agree prevents Agree in the feature [+Part] over a DP
bearing [–Part], correctly ruling out hierarchy-violating structures like (4b) above, shown
in (14).1 In a grammatical “plateau” configuration with two [+Part] DPs (§3), the agreeing
probe can target features of both DPs without violating (13), as in (15). Finally, in hierarchyobeying configurations (i.e., Part > 3, in (4a)), the verbal head agrees with DP1 in [+Part]
and stops; since [–Part] features require no licensing (12), ungrammaticality does not arise.
(14) * 3 > Participant:
[ Probe0 [ DP[– PART ] . . . [ . . . DP[+ PART ] ]]]
5
(15)
Participant > Participant:
[ Probe0 [ DP[+ PART ] . . . [ . . . DP[+ PART ] ]]]
(16)
Participant > 3:
[ Probe0 [ DP[+ PART ] . . . [ . . . DP[– PART ] ]]]
This system accounts for both weak PCC effects and German copula constructions—both
configurations in which two DPs are in the domain of a single agreeing probe. It furthermore
provides a rationale for why no such hierarchy effect arises in English. In German, both
DP1 and DP2 bear nominative case, and nominative DPs are accessible agreement targets
in German. In English, DP2 (the predicate nominal) is marked with accusative case and
hence invisible to the agreeing verbal head (Bobaljik 2008). Because the PLC requires only
accessible DPs to be licensed, no violation arises (see discussion in Preminger 2016).
We now turn to the effects of the number hierarchy from (5) above. We saw that
German copula constructions disallow SG > PL configurations while tolerating all other
combinations. A similar PCC-style account of this asymmetry becomes available if the
feature calculus and licensing conditions above are extended to number features, as in (17).
See Anagnostopoulou 2003, Rezac 2008, and Baker 2011 for similar conditions.
(17)
Feature Licensing Condition: A marked feature F on a DP that is a viable agreement
target (as far as its case, etc. is concerned), and for which there is a clausemate F
probe, must participate in a valuation relation.
Under the assumption that [+Pl(ural)] is the marked number specification, and the φ-probe
may license number features, then SG > PL configurations are immediately ruled out as
violations of FLC, just as in the person configurations above. As was the case with person,
1 We
assume that both DPs in the German copula construction are generated internally to a small clause,
and that the agreeing Probe schematized in (14)–(16) is finite T0 .
Coon, Keine, & Wagner
this hierarchy effect is limited to copula constructions in German because it is only in these
constructions that we find two nominative DPs, both visible to the agreeing probe.
4.2
The absence of “Number Case Constraints”
Our account of the number hierarchy effects raises an interesting question. While person
restrictions in ditransitive constructions are well-documented, there do not appear to be
similar restrictions with respect to number features. In other words, there is no “Number
Case Constraint” in double-object clitic configurations (Nevins 2011). In light of the German
facts and our account, it is not a priori clear why number does not induce hierarchy effects
in ditransitive structures, as it does in copulas.2
We propose that the crucial difference between German copula constructions and PCCinducing ditransitive structures is that only the latter induce clitic doubling. More concretely,
our account of this difference relies on three independently-motivated proposals. First, we
adopt the proposal that person and number are separate probes (e.g., Béjar & Rezac 2003),
π0 and #0 , respectively. Second, we assume further that #0 is universally located higher in
the tree than π0 (Béjar & Rezac 2003, Preminger 2011) so that π0 will always probe first.
Finally, we adopt proposals of Anagnostopoulou (2003) and Preminger (2009) that clitic
doubling renders the doubled DP invisible to subsequent operations.
As a consequence, in ditransitive constructions, clitic doubling of an indirect object
as a result of Agree with π0 removes it as an intervener, clearing the way for subsequent
Agree between #0 and the direct object. As a consequence, if clitic doubling takes place, the
indirect object will never cause intervention for number agreement with the direct object,
deriving the absence of “Number Case Constraint” effects, as shown in (18).
(18)
[vP #0 [ π0 [ApplP DPIO [VP DPDO ]]]]
= Ditransitive PCC
clitic-double
It is a general property of German that it lacks clitic doubling. Thus, Agree between π0
and DP1 in copula constructions does not render DP1 invisible for subsequent Agree. DP1
therefore still incurs intervention for Agree between #0 and DP2 if Contiguous Agree is
violated (19). Number is hence subject to the same intervention effects as person in German.
(19)
[TP #0 [ π0 [PredP DPSUBJ [ DPPRED ]]]]
2 Nevins
= German copula
(2011) attributes the difference in behavior between person and number to an ontological
difference between the two types of features: person features are binary, while number features are privative.
Thus, while 3rd person contains a negative feature specification, singular number corresponds to the absence
of a feature. Notice, however, that one of the key arguments Nevins (2007) makes for 3rd person bearing a
person specification applies to number as well: In English, 3rd person singular verb agreement is expressed
with -s, a vocabulary item that must consequently be specified for both 3rd person and singular number. Nevins
(2007) concludes from this that 3rd person cannot simply be the absence of a person feature. By the same
reasoning, singular number cannot be the absence of a number feature either, contra Nevins (2011). See also
Béjar 2011 (and work cited there), as well as Preminger 2014, for other problems with Nevins’ account.
Hierarchy effects in copular constructions
5.
Summary and future work
In this paper, we presented a new take on agreement patterns in German copula constructions.
On the maximize agreement account, agreement in German copulas must target the featurally
most specified DP. Under the alternative proposed here, agreement in German consistently
targets the subject; the appearance of non-subject agreement—as in (2) above—is the result
of syntactic inversion of the predicate nominal above the subject (see e.g. Heggie 1988 and
discussion in B&K). On both accounts, the corresponding effect does not arise in English,
because the predicate nominal is accusative, and thus inaccessible to an agreeing probe.
New to our account is the finding that certain subject–predicate combinations are simply
ineffable in semantically asymmetric so-called ‘assumed identity sentences’. While Heycock
(2012) reports equivalent sentences to be grammatical under various feature combinations,
we presented evidence for hierarchy effects. We adopted components of existing PCC
accounts in order to model the German effects. Specifically, our proposal made use of the
Person Licensing Condition, requiring [+Part] features on accessible DPs to be licensed by
an agreeing probe. Person hierarchy effects are thus correctly expected to appear when there
is a many-to-one relation between accessible DPs and the agreeing probe.
In the last section, we addressed the question of why number effects appear in German
copula constructions, despite the fact that there is no constraint restricting combinations
of direct and indirect object clitics based on their number features (i.e. no “Num-CC”).
Our account relied on the independently-motivated proposal that clitic-doubling renders
the doubled DP invisible to future operations. Assuming that the φ-probe is complex, and
that person (π0 ) probes before number (#0 ), the number probe will always have access to
the direct object (DP2 ), since π0 will remove the indirect object (DP1 ) as an intervener.
However, since no clitic-doubling is present in German copula constructions, the subject
DP1 will intervene between the probe and the lower DP2 . This proposal explains the absence
of Num-CC effects in ditransitives, without resorting to ontological differences in the nature
of person and number features (contra Nevins 2011).
Finally, our account makes a number of testable predictions. Specifically, all else being
equal, we expect to find hierarchy effects in copula constructions in other languages, so
long as (i) both DPs in predicate nominals are accessible to agreement, and (ii) there is a
clausemate agreeing probe (see Preminger 2016). The appearance of special agreement in
copula constructions has been described in a number of languages (see Heycock 2012 and
K&P, i.a.), and it remains to be seen whether our approach can be applied to these cases
as well. Furthermore, we predict that both person and number hierarchy effects may be
present, unless clitic-doubling is involved. Since PCC constructions involve clitic-doubling
by definition (i.e. it is not the combinations of arguments that are ruled out, but combinations
of clitic-doubles), these are correctly expected to lack number effects.
References
Adger, David, & Daniel Harbour. 2007. Syntax and Syncretisms of the Person Case
Constraint. Syntax 10:2–37.
Coon, Keine, & Wagner
Anagnostopoulou, Elena. 2003. The syntax of ditransitives: Evidence from clitics. Berlin:
Mouton de Gruyter.
Anagnostopoulou, Elena. 2005. Strong and Weak Person Restrictions: A feature checking
analysis. In Clitic and Affix Combinations: Theoretical Perspectives, ed. Lorie Heggie
& Francisco Ordóñez, Linguistics Today, 199–235. Amsterdam: John Benjamins.
Baker, Mark C. 2011. When agreement is for number and gender but not person. Natural
Language and Linguistic Theory 29:875–915.
Béjar, Susana. 2011. Remarks on omnivory and complementarity: A commentary on the
paper by Andrew Nevins. Natural Language and Linguistic Theory 29:973–997.
Béjar, Susana, & Arsalan Kahnemuyipour. to appear. Non-canonical agreement in copular
clauses. Journal of Linguistics .
Béjar, Susana, & Milan Rezac. 2003. Person Licensing and the Derivation of PCC Effects.
In Romance Linguistics: Theory and Acquisition, ed. Ana Teresa Perez-Leroux & Yves
Roberge, 49–62. Amsterdam: John Benjamins.
Béjar, Susana, & Milan Rezac. 2009. Cyclic Agree. Linguistic Inquiry 40:35–73.
Bobaljik, Jonathan David. 2008. Where’s Phi? Agreement as a Post-Syntactic Operation.
In Phi Theory: Phi-features across interfaces and modules, ed. Daniel Harbour, David
Adger, & Susana Béjar, 295–328. Oxford: Oxford University Press.
Christensen, Rune Haubo B. 2015. ordinal – regression models for ordinal data. r package
version 2015.6-28. URL http://www.cran.r-project.org/package=ordinal.
Heggie, Lorie A. 1988. The syntax of copular structures. Doctoral dissertation, University
of Southern California, Los Angeles, CA.
Heycock, Caroline. 2012. Specification, equation, and agreement in copular sentences.
Canadian Journal of Linguistics 57:209–240.
Hiraiwa, Ken. 2001. Multiple agree and the defective intervention constraint in Japanese.
In The proceedings of the MIT-Harvard joint conference (HUMIT 2000), ed. Ora Matushansky, MIT Working Papers in Linguistics 40, 67–80. Cambidge, MA: MITWPL.
Nevins, Andrew. 2007. The representation of third person and its consequences for PersonCase effects. Natural Language and Linguistic Theory 25:273–313.
Nevins, Andrew. 2011. Mulitple agree with clitics: person complementarity vs. omnivorous
number. Natural Language and Linguistic Theory 939–971.
Preminger, Omer. 2009. Breaking Agreements: Distinguishing Agreement and CliticDoubling by Their Failures. Linguistic Inquiry 40:619–666.
Preminger, Omer. 2011. Asymmetries between person and number in syntax: A commentary
on Baker’s SCOPA. Natural Language and Linguistic Theory 29:917–937.
Preminger, Omer. 2014. Agreement and its failures. Cambridge, MA: MIT Press.
Preminger, Omer. 2016. What the PCC tells us about “abstract” agreement, head movement,
and locality. Ms. University of Maryland.
Rezac, Milan. 2008. The syntax of eccentric agreement: The Person Case Constraint
and Absolutive Displacement in Basque. Natural Language and Linguistic Theory
26:61–106.
Jessica Coon, Stefan Keine, Michael Wagner
[email protected], [email protected], [email protected]